OPTIMUM WEIGHT SMOOTHING IN FINITE DATA

In this correspondence, we consider a generalized smoothing problem an d develop a procedure to obtain a set of optimum weights which gives m inimum mean-squared error (MSE) in the estimates of directions of arri val of signals in finite data when the signals are arbitrarily correla ted. Using the optimum weights, we study the optimum tradeoff between the number of subarrays and the subarray size for a fixed total size o f the array. The computation of optimum weights, however, requires ful l knowledge of the scenario. Since exact DOA's, powers, and correlatio ns of signals are unknown a priori, we give a method to estimate these weights from the observed finite data. We also show through empirical studies that the optimum weights can be approximated with Taylor weig hts which serve as near-optimum weights. Simulation results are includ ed to support the theoretical assertions.